forked from dmlc/xgboost
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_eval_metrics.py
241 lines (199 loc) · 8.74 KB
/
test_eval_metrics.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
import xgboost as xgb
import testing as tm
import numpy as np
import pytest
rng = np.random.RandomState(1337)
class TestEvalMetrics:
xgb_params_01 = {
'verbosity': 0,
'nthread': 1,
'eval_metric': 'error'
}
xgb_params_02 = {
'verbosity': 0,
'nthread': 1,
'eval_metric': ['error']
}
xgb_params_03 = {
'verbosity': 0,
'nthread': 1,
'eval_metric': ['rmse', 'error']
}
xgb_params_04 = {
'verbosity': 0,
'nthread': 1,
'eval_metric': ['error', 'rmse']
}
def evalerror_01(self, preds, dtrain):
labels = dtrain.get_label()
return 'error', float(sum(labels != (preds > 0.0))) / len(labels)
def evalerror_02(self, preds, dtrain):
labels = dtrain.get_label()
return [('error', float(sum(labels != (preds > 0.0))) / len(labels))]
@pytest.mark.skipif(**tm.no_sklearn())
def evalerror_03(self, preds, dtrain):
from sklearn.metrics import mean_squared_error
labels = dtrain.get_label()
return [('rmse', mean_squared_error(labels, preds)),
('error', float(sum(labels != (preds > 0.0))) / len(labels))]
@pytest.mark.skipif(**tm.no_sklearn())
def evalerror_04(self, preds, dtrain):
from sklearn.metrics import mean_squared_error
labels = dtrain.get_label()
return [('error', float(sum(labels != (preds > 0.0))) / len(labels)),
('rmse', mean_squared_error(labels, preds))]
@pytest.mark.skipif(**tm.no_sklearn())
def test_eval_metrics(self):
try:
from sklearn.model_selection import train_test_split
except ImportError:
from sklearn.cross_validation import train_test_split
from sklearn.datasets import load_digits
digits = load_digits(n_class=2)
X = digits['data']
y = digits['target']
Xt, Xv, yt, yv = train_test_split(X, y, test_size=0.2, random_state=0)
dtrain = xgb.DMatrix(Xt, label=yt)
dvalid = xgb.DMatrix(Xv, label=yv)
watchlist = [(dtrain, 'train'), (dvalid, 'val')]
gbdt_01 = xgb.train(self.xgb_params_01, dtrain, num_boost_round=10)
gbdt_02 = xgb.train(self.xgb_params_02, dtrain, num_boost_round=10)
gbdt_03 = xgb.train(self.xgb_params_03, dtrain, num_boost_round=10)
assert gbdt_01.predict(dvalid)[0] == gbdt_02.predict(dvalid)[0]
assert gbdt_01.predict(dvalid)[0] == gbdt_03.predict(dvalid)[0]
gbdt_01 = xgb.train(self.xgb_params_01, dtrain, 10, watchlist,
early_stopping_rounds=2)
gbdt_02 = xgb.train(self.xgb_params_02, dtrain, 10, watchlist,
early_stopping_rounds=2)
gbdt_03 = xgb.train(self.xgb_params_03, dtrain, 10, watchlist,
early_stopping_rounds=2)
gbdt_04 = xgb.train(self.xgb_params_04, dtrain, 10, watchlist,
early_stopping_rounds=2)
assert gbdt_01.predict(dvalid)[0] == gbdt_02.predict(dvalid)[0]
assert gbdt_01.predict(dvalid)[0] == gbdt_03.predict(dvalid)[0]
assert gbdt_03.predict(dvalid)[0] != gbdt_04.predict(dvalid)[0]
gbdt_01 = xgb.train(self.xgb_params_01, dtrain, 10, watchlist,
early_stopping_rounds=2, feval=self.evalerror_01)
gbdt_02 = xgb.train(self.xgb_params_02, dtrain, 10, watchlist,
early_stopping_rounds=2, feval=self.evalerror_02)
gbdt_03 = xgb.train(self.xgb_params_03, dtrain, 10, watchlist,
early_stopping_rounds=2, feval=self.evalerror_03)
gbdt_04 = xgb.train(self.xgb_params_04, dtrain, 10, watchlist,
early_stopping_rounds=2, feval=self.evalerror_04)
assert gbdt_01.predict(dvalid)[0] == gbdt_02.predict(dvalid)[0]
assert gbdt_01.predict(dvalid)[0] == gbdt_03.predict(dvalid)[0]
assert gbdt_03.predict(dvalid)[0] != gbdt_04.predict(dvalid)[0]
@pytest.mark.skipif(**tm.no_sklearn())
def test_gamma_deviance(self):
from sklearn.metrics import mean_gamma_deviance
rng = np.random.RandomState(1994)
n_samples = 100
n_features = 30
X = rng.randn(n_samples, n_features)
y = rng.randn(n_samples)
y = y - y.min() * 100
reg = xgb.XGBRegressor(tree_method="hist", objective="reg:gamma", n_estimators=10)
reg.fit(X, y, eval_metric="gamma-deviance")
booster = reg.get_booster()
score = reg.predict(X)
gamma_dev = float(booster.eval(xgb.DMatrix(X, y)).split(":")[1].split(":")[0])
skl_gamma_dev = mean_gamma_deviance(y, score)
np.testing.assert_allclose(gamma_dev, skl_gamma_dev, rtol=1e-6)
@pytest.mark.skipif(**tm.no_sklearn())
def test_gamma_lik(self) -> None:
import scipy.stats as stats
rng = np.random.default_rng(1994)
n_samples = 32
n_features = 10
X = rng.normal(0, 1, size=n_samples * n_features).reshape((n_samples, n_features))
alpha, loc, beta = 5.0, 11.1, 22
y = stats.gamma.rvs(alpha, loc=loc, scale=beta, size=n_samples, random_state=rng)
reg = xgb.XGBRegressor(tree_method="hist", objective="reg:gamma", n_estimators=64)
reg.fit(X, y, eval_metric="gamma-nloglik", eval_set=[(X, y)])
score = reg.predict(X)
booster = reg.get_booster()
nloglik = float(booster.eval(xgb.DMatrix(X, y)).split(":")[1].split(":")[0])
# \beta_i = - (1 / \theta_i a)
# where \theta_i is the canonical parameter
# XGBoost uses the canonical link function of gamma in evaluation function.
# so \theta = - (1.0 / y)
# dispersion is hardcoded as 1.0, so shape (a in scipy parameter) is also 1.0
beta = - (1.0 / (- (1.0 / y))) # == y
nloglik_stats = -stats.gamma.logpdf(score, a=1.0, scale=beta)
np.testing.assert_allclose(nloglik, np.mean(nloglik_stats), rtol=1e-3)
def run_roc_auc_binary(self, tree_method, n_samples):
import numpy as np
from sklearn.datasets import make_classification
from sklearn.metrics import roc_auc_score
rng = np.random.RandomState(1994)
n_samples = n_samples
n_features = 10
X, y = make_classification(
n_samples,
n_features,
n_informative=n_features,
n_redundant=0,
random_state=rng
)
Xy = xgb.DMatrix(X, y)
booster = xgb.train(
{
"tree_method": tree_method,
"eval_metric": "auc",
"objective": "binary:logistic",
},
Xy,
num_boost_round=8,
)
score = booster.predict(Xy)
skl_auc = roc_auc_score(y, score)
auc = float(booster.eval(Xy).split(":")[1])
np.testing.assert_allclose(skl_auc, auc, rtol=1e-6)
X = rng.randn(*X.shape)
score = booster.predict(xgb.DMatrix(X))
skl_auc = roc_auc_score(y, score)
auc = float(booster.eval(xgb.DMatrix(X, y)).split(":")[1])
np.testing.assert_allclose(skl_auc, auc, rtol=1e-6)
@pytest.mark.skipif(**tm.no_sklearn())
@pytest.mark.parametrize("n_samples", [4, 100, 1000])
def test_roc_auc(self, n_samples):
self.run_roc_auc_binary("hist", n_samples)
def run_roc_auc_multi(self, tree_method, n_samples):
import numpy as np
from sklearn.datasets import make_classification
from sklearn.metrics import roc_auc_score
rng = np.random.RandomState(1994)
n_samples = n_samples
n_features = 10
n_classes = 4
X, y = make_classification(
n_samples,
n_features,
n_informative=n_features,
n_redundant=0,
n_classes=n_classes,
random_state=rng
)
Xy = xgb.DMatrix(X, y)
booster = xgb.train(
{
"tree_method": tree_method,
"eval_metric": "auc",
"objective": "multi:softprob",
"num_class": n_classes,
},
Xy,
num_boost_round=8,
)
score = booster.predict(Xy)
skl_auc = roc_auc_score(y, score, average="weighted", multi_class="ovr")
auc = float(booster.eval(Xy).split(":")[1])
np.testing.assert_allclose(skl_auc, auc, rtol=1e-6)
X = rng.randn(*X.shape)
score = booster.predict(xgb.DMatrix(X))
skl_auc = roc_auc_score(y, score, average="weighted", multi_class="ovr")
auc = float(booster.eval(xgb.DMatrix(X, y)).split(":")[1])
np.testing.assert_allclose(skl_auc, auc, rtol=1e-6)
@pytest.mark.parametrize("n_samples", [4, 100, 1000])
def test_roc_auc_multi(self, n_samples):
self.run_roc_auc_multi("hist", n_samples)